MVP safety approach · 11 August 2026
Safety without silencing
Strong feelings stay expressible. Operational harm instructions do not.
Difficult words are welcome
Anger, grief, profanity, resentment, fear, shame, personal confessions, and statements such as “I hate you” are part of what Let It Leave is for. They are not blocked merely because they are intense, vulgar, or uncomfortable.
What is limited
Let It Leave will not help turn requests for step-by-step instructions to kill, injure, poison, build weapons, or exploit children into a generated visual. When possible, it asks you to reframe the text as the feeling or burden you want to release.
Immediate-risk language
If the text suggests an immediate intention to harm yourself or someone else, Let It Leave avoids sending it to the image model. It can preserve the words as a private-feeling unsent letter and displays a prompt to contact local emergency services or a trusted person. This is a safety response, not a diagnosis.
No hidden reporting
Let It Leave does not automatically send your prompt to police, employers, friends, or an analytics dashboard. The MVP does not maintain an account-linked prompt history. Network and hosting providers may still process limited technical logs needed to deliver and secure the service.
Not crisis care
Let It Leave is a reflective visual ritual, not therapy, crisis intervention, or medical advice. If you might act now, contact local emergency services and a person who can stay with you. In Romania and across the EU, the emergency number is 112.
Review before launch
The current moderation is intentionally narrow and rules-based for the MVP. Before a broad launch it should be tested with multilingual safety evaluations, reviewed by a qualified safety professional, and paired with region-appropriate support information.